CN106650608A - Identification method for rectangle locating frame in test paper without locating points - Google Patents
Identification method for rectangle locating frame in test paper without locating points Download PDFInfo
- Publication number
- CN106650608A CN106650608A CN201610938847.3A CN201610938847A CN106650608A CN 106650608 A CN106650608 A CN 106650608A CN 201610938847 A CN201610938847 A CN 201610938847A CN 106650608 A CN106650608 A CN 106650608A
- Authority
- CN
- China
- Prior art keywords
- line segment
- rectangle
- line
- paper
- error
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an identification method for a rectangle locating frame of to-be-processed test paper in test paper without locating points. The rectangle frame can be accurately located. According to conventional methods, rectangle identification is generally performed by searching for a rectangle profile and a straight line crossing point; in the first method, an error is generated when multiple parts of the rectangle are broken or line segments are adhered on the rectangle frame, so that the robustness is low; and the second method is disturbed by straight line identification, and judgment is difficultly carried out when the line segments are identified. According to the method provided by the invention, the line segments detected by Hough transform are preprocessed, and weight calculation is added, so that the problems of rectangle breaking, pixel disturbance and multiple line segments can be well solved and the rectangle can be accurately located and identified.
Description
Technical field
The present invention relates to image recognition positioning field, it is more particularly to a kind of based on Hough transformation without anchor point paper rectangle
The recognition methods of posting.
Background technology
With the increasingly specification improved with examination system of social education system, the type and mode of examination also constantly increase
Many but traditional papery examinations remain the main flow of current examination.Current automatic marking papers system mainly have optical character reader (OCR) and
Marking system based on image procossing.And answer paper is roughly divided into two kinds, a kind of have a site, and one kind is without anchor point
's；For currently without the paper of anchor point, the identification of the method for positioning mainly by question number and multiple-choice question full-filling region is common
Determine the anchor point of paper, this method needs maturation, accurately recognizes storehouse, and commercial cost is higher；The present invention proposes one kind
Based on the rectangle recognizer of Hough transformation, the rectangle frame in paper, the rectangle fracture of antagonism paper, mould can be relatively recognized accurately
Various situations such as paste, inclination, robustness is preferable.
The content of the invention
It is an object of the invention to overcome the shortcoming of prior art with it is not enough, there is provided it is a kind of accurate, stable without anchor point
Paper rectangle frame recognition methods.
The rectangle posting recognition methods of pending paper, comprises the steps of in a kind of paper without anchor point：
1) paper is read in, carries out binaryzation, correction and process；
2) rectangular area information of template record is read, with " largest contours searching " method the most bull wheel in region is obtained
Exterior feature, is fitted with minimum rectangle to profile；
3) judge whether the rectangle size for fitting is consistent with the rectangle size of record, method is one error model of setting
Enclose, be respectively compared the length of two rectangles and the error of height whether in setting range；If in setting range, the fitting
Rectangle out is the rectangle for needing identification；If not in setting range, carrying out step 4)；
4) line segment of area information is detected with Hough transformation, and the line segment to detecting is pre-processed, including classification,
Merge；
5) sequence is weighted to pretreated line segment；After being calculated the weights of every line segment, according to weights from big
Line segment is ranked up to little；
6) out, remaining gives up the line segments extraction that weights are located at into front three；
7) respectively take out a line segment from four direction up and down to be fitted, the method for fitting is：Obtain line segment with
The intersection point Point2 of the intersection point Point1 of left line segment, lower line segment and right line segment, being fitted the rectangle length for obtaining is：Point2.x-
Point1.x；Highly it is：Point2.y-Point1.y；
8) compare the rectangle recorded in template to grow tall, if error is within setting range, then it is assumed that recognize successfully, return
The coordinate of fitted rectangle and grow tall, it is the rectangle for needing identification that this fits the rectangle for coming；If the error of length or height its
In have one exceed setting range, then carry out next step；
9) next candidate line sections are taken out in the line segment aggregate of top, is fitted again；
10) when top line segment aggregate is all completed and does not find suitable rectangle, by the line segment of the direction
Index resets, and the authority for changing line segment is moved to next direction by the order according to " upper, left, down, right "；
11) after all line segment aggregates whole test fitting on four direction is finished, if suitable rectangle is not found
Output " rectangle does not find ".
Further, line segment weights include segment positions error weights and line segment length error weights；
Segment positions error weight computing formula is：
Site error weights=0.8^abs (current line segment sequence number-record line segment sequence number)；
Line segment length error weight computing formula is：
Error in length weights=(line segment length-minimum line segment length)/(maximum line segment length-minimum line segment length).
The selection rectangle method of blank template paper, comprises the steps of in a kind of paper without anchor point：
1) binary conversion treatment is carried out by scanning the original paper image for obtaining；
2) correction process is carried out to it；
3) after the image after being corrected, three maximum rectangles of area on paper are found out with " searching largest contours " method,
And record its position；
4) rectangle is carried out once " judgement of rectangular line overlap ", it is ensured that largest contours are searched out intending with straight line for rectangle
The rectangle size closed out is consistent.
Further, correction is processed to be specially and searches out the maximum profile of area in image, then with the square of a minimum
Shape surrounds profile, calculates the angle of rectangle, draws the angle of inclination of paper, then paper is rectified a deviation.
Further, " judgement of rectangular line overlap " method is as follows：
1) the pixel width of rectangle frame is determined, general 3～6 pixels are rule of thumb set to 5；
2) from positioning region from top to bottom transversal scanning image, 5 pixels are scanned successively downwards, if wherein there is black picture
Vegetarian refreshments, then stop scanning, and adds 1 by total pixel SumPix of the row；
3) after a line is scanned through, the size of SumPix and positioning area field width is compared, if SumPix is more than positioning region
Wide 80%, then it is assumed that the row is likely to be the length of side of rectangle, and stops scanning；Otherwise, then by the row pixel whole zero setting；
4) from upper and lower, left and right four direction it is purged operation to image successively, filters off noise as much as possible；
5) straight-line detection is carried out with Hough transformation function to the rectangular region image after removing is processed, obtains the area
The line segment information in domain；
6) line segment to detecting carries out after-treatment；
7) serial number information of rectangle line segment on four direction is recorded.
Further, the after-treatment step is as follows：
1) line segment is divided into four classes, one is the line segment on the vertical left side, and two is the line segment on vertical the right, and three is above level
Line segment, four is the line segment of horizontal bottom, sorting technique：According to line segment coordinate calculate line segment angle, then compare line segment coordinate with
The magnitude relationship of regional center, draws the classification of line segment；
2) line segment is merged：Two lines section is taken out from sorted line segment, their distances on respective direction are calculated, is hung down
Straight line segment distance is the absolute value for subtracting each other of y-axis coordinate, and the line segment distance of level is the absolute value for subtracting each other of x-axis coordinate, than
Whether compared with distance in 5 pixels, if then merging in two straight lines, the method for merging is the seat for taking the respective end points of two straight lines
Mark；
3) line segment after merging is sorted, because to record index sequence of the rectangular shaped rim line segment in all line segments for identifying
Number, it is therefore desirable to the line segment of diverse location is needed to make different sequences：
Top line segment：Sorted according to y-axis size；
Bottom line segment：Sort according to sections bottom positional distance size is found；
Left side line segment：Sorted according to x-axis size；
Right side line segment：Sort apart from size according to region right positions are found.
The present invention has the following advantages and effect relative to prior art：Rectangle frame can be accurately positioned, relative to
Method in the past, rectangle identification is general with finding rectangular profile and finding straight-line intersection, and first method ruptures in rectangle many places
Or will malfunction when having line segment to stick on rectangle frame, robustness is not high；Second method can be disturbed by Straight Line Identification, and
And be difficult to judge in the case where a plurality of line segment is identified, and the method in the present invention, the line segment of Hough transformation detection is carried out
Pretreatment, and weight computing is added, rectangle fracture, pixel interference, a plurality of line segment can be well solved the problems, such as, it is accurately right
Rectangle carries out fixation and recognition.
Description of the drawings
Fig. 1 is that Hough change formula graphic represents figure；
Fig. 2 is the selection rectangle flow chart on blank template paper；
Fig. 3 is the flow chart that rectangle is found on test sample paper.
Specific embodiment
Below in conjunction with accompanying drawing, the present invention is described in detail：
Based on Hough transformation, Hough transformation is one of the classical means in image conversion to the present invention, be mainly used to from
The geometry (e.g., straight line, circle etc.) with certain same characteristic features is isolated in image.Hough transformation finds the side of straight line and circle
Method is compared can preferably reduce noise jamming with other methods.Classical Hough transformation is commonly used to detection of straight lines, circle, oval
Deng.
Hough transformation detection of straight lines algorithm idea：
Each pixel coordinate point becomes the contributive unified metric of Chengdu straight line speciality, a simply example through conversion
It is as follows：Straight line is in the picture the set of series of discrete point, by the discrete polar coordinates formula of a straight line, can be with table
The discrete point geometric equality for reaching straight line is as follows：
X*cos (theta)+y*sin (theta)=r；
Wherein angle theta refers to the angle between r and X-axis, and r is to rectilinear geometry vertical range.It is any in Points on Straight Line,
X, y can be expressed, wherein r, and theta is constant.The formula graphic represents as shown in Figure 1.
But in the image processing field realized, pixel coordinate P (x, y) of image is known, and r, theta are then us
The variable to be found.If we can draw each (r, theta) value according to pixel point coordinates P (x, y) value, then just
Polar coordinates hough space system is transformed into from image cartesian coordinate system, this change from point to curve is referred to as straight line suddenly
Husband converts.Conversion is by quantifying Hough parameter space for limited value interval decile or cumulative grid.When Hough transformation algorithm
Start, each pixel coordinate point P (x, y) is switched to above the curve point of (r, theta), is added to corresponding grid data
Point, when a crest occurs, illustrates with the presence of straight line.
The algorithm of rectangle identification is divided into two big steps, and one is the record of rectangular information in template, and two is square on pending paper
The fixation and recognition of shape.Its overall thought is that rectangle size, frame line segment are first recorded on template paper in owning for identifying
Ranking index and length in straight line, according to these information for recording the identification of rectangle is made on pending paper.Identification
Method is divided to two kinds, and one is that largest contours are found, this method fast operation, the feelings at rectangle ruptures very big but only
Stand good under condition, even if recognition effect is still good but larger by pixel interference in the case of rectangle is fuzzy, the font on paper
If exceeding rectangle, the failure of identification is may result in；Two is that the fitting a straight line detected according to Hough transformation goes out one with record letter
The most close rectangle of breath, this method can well resist the situation of rectangle many places fracture, and be disturbed less by written word soma,
But the speed of service can be relatively slow.Comprehensive two methods can preferably identify the rectangle of most paper.
First, the selection rectangle method of blank template paper：
1) may be coloured by scanning the original paper image for obtaining, grey, in order to be uniformly processed, also for letter
Change operation, therefore gray processing, binary conversion treatment are carried out firstly the need of to image.Binaryzation adopts experience binary processing method.
2) original image there may be inclined problem, and also have after binaryzation carries out correction process to it.Entangling here
Folk prescription method adopts " maximum rectangle correction " method.Concrete grammar is：The maximum profile of area in image is searched out, then with one most
Little rectangle surrounds profile, calculates the angle of rectangle, draws the angle of inclination of paper, then paper is rectified a deviation.
3) after the image after being corrected, three maximum rectangles of area on paper are found out with " searching largest contours " method,
And record its position.
4) three rectangles for obtaining might not can serve as being set to rectangle, because the reason for frame ruptures, may
The rectangle for searching out is caused to diminish, it is therefore desirable to carry out once to rectangle " judgement of rectangular line overlap ", effect is to ensure that most
Big profile search out come rectangle the rectangle size with fitting a straight line out it is consistent.
" judgement of rectangular line overlap " method is as follows：
Word in paper can affect the identification of line segment, in order to filter out the pixel of interference as far as possible, it is necessary first to right
Rectangular area is purged process, and step is as follows：
1) the pixel width of rectangle frame is determined, general 3～6 pixels are rule of thumb set to 5；
2) from positioning region from top to bottom transversal scanning image, 5 pixels are scanned successively downwards, if wherein there is black picture
Vegetarian refreshments, then stop scanning, and adds 1 by total pixel SumPix of the row；
3) after a line is scanned through, the size of SumPix and positioning area field width is compared, if SumPix is more than positioning region
Wide 80%, then it is assumed that the row is likely to be the length of side of rectangle, and stops scanning；Otherwise, then by the row pixel whole zero setting.
4) from upper and lower, left and right four direction it is purged operation to image successively, filters off noise as much as possible；
5) straight-line detection is carried out with Hough transformation function to the rectangular region image after removing is processed, obtains the area
The line segment information in domain
6) because line Segment has the reasons such as thickness, fracture, the general effect of the line segment that Hough transformation is detected is not very
Perfection, such as one line segment will detect that a plurality of, it is therefore desirable to which the line segment to detecting carries out after-treatment.
7) serial number information of rectangle line segment on four direction is recorded.
Wherein, after-treatment process step is as follows：
1. line segment is divided into four classes, one is the line segment on the vertical left side, and two is the line segment on vertical the right, and three is above level
Line segment, four is the line segment of horizontal bottom, sorting technique：According to line segment coordinate calculate line segment angle, then compare line segment coordinate with
The magnitude relationship of regional center, draws the classification of line segment
2. line segment is merged.Two lines section is taken out from sorted line segment, their distances on respective direction are calculated, is hung down
Straight line segment distance is the absolute value for subtracting each other of y-axis coordinate, and the line segment distance of level is the absolute value for subtracting each other of x-axis coordinate, than
Whether compared with distance in 5 pixels, if then merging in two straight lines, the method for merging is the seat for taking the respective end points of two straight lines
Mark.
3. the line segment after merging is sorted, because to record index sequence of the rectangular shaped rim line segment in all line segments for identifying
Number, it is therefore desirable to the line segment of diverse location is needed to make different sequences：
Top line segment：Sorted according to y-axis size；
Bottom line segment：Sort according to sections bottom positional distance size is found；
Left side line segment：Sorted according to x-axis size；
Right side line segment：Sort apart from size according to region right positions are found；
2nd, the rectangle positioning identifying method on pending paper：
1) paper is read in, carries out binaryzation, correction and process.
2) rectangular area information of template record is read, with " largest contours searching " method the most bull wheel in region is obtained
Exterior feature, is fitted with minimum rectangle to profile.
3) judge whether the rectangle size for fitting is consistent with the rectangle size of record, method is one error model of setting
Enclose, here according to experience 8 pixels are set to, whether be respectively compared the length of two rectangles and the error of height in setting range
It is interior；If in setting range, it is the rectangle for needing identification that this fits the rectangle for coming.If both errors are larger, carry out
Next step.
4) line segment of area information is detected with Hough transformation, and the line segment to detecting is pre-processed, including classification,
Merge；
5) sequence is weighted to pretreated line segment.Weighted Segments sequence is based on such a problem：Hough
Conversion will detect that many bar line segments, and on earth which bar line segment is only the sideline of rectangleIf all of line segment is all tested into one
All over so as to if the rectangle for drawing optimum, it is assumed that there are 10 line segments in each direction, then fit the plan required for optimum rectangle
Close and attempt at most being likely to be：10^4=10000 time.The efficiency of so identification will be very low, it is therefore desirable to line segment is carried out to add
Power sequence, filter out weights highest, that is, the line segment being most likely to be on rectangle sideline, on the contrary it will not be possible to line segment abandon, most
After obtain three line segments of weights highest, then every time fitted rectangle worst also simply 3^4=81 time, and in many cases,
Because having filtered out weights highest line segment, therefore typically only needed to the trial of 3-5 time just can obtain the rectangle of optimum.
So Weighted Segments sequence is a very important step.Line segment weights include the content of two aspects, and one is segment positions error
Weights, two is line segment length error weights.Because originally have recorded the sideline sequence number of rectangle in all directions in template,
Therefore the line segment closer to the sequence number is more likely to be the sideline of rectangle, but because the expansion of template paper and paper to be identified is known
The reason for other region is not quite identical, it is exactly not necessarily the sideline of rectangle that sequence number is identical, so also to add sentencing for line segment length
Disconnected, length is more likely to be the rectangle sideline for needing to find closer to the line segment of rectangle side line length in template.
Segment positions error weights determine that its computing formula is by the sequence number difference of the line segment sequence number after sorting and record：
Site error weights=0.8^abs (current line segment sequence number-record line segment sequence number)；
Formula Parsing：When sequence number error is 0, weights are 1, and when sequence number error increases, weights are due to due to power operation
It is rapid to reduce, significantly the line segment of each stratum can be made a distinction.
Line segment length error weights need to be normalized error in order to avoid the excessive impact of error information,
Normalized method is min-max standardization.The rectangle side line length error of every line segment and record is calculated first, and preserves mistake
Poor maximin, be by the normalized formula of error：
Error in length weights=(line segment length-minimum line segment length)/(maximum line segment length-minimum line segment length)；
After normalization, each error information can play identical action effect.
After being calculated the weights of every line segment, according to weights line segment is ranked up from big to small.
6) out, remaining gives up the line segments extraction that weights are located at into front three.
7) respectively take out a line segment from four direction up and down to be fitted, the method for fitting is：Obtain line segment with
The intersection point Point2 of the intersection point Point1 of left line segment, lower line segment and right line segment, being fitted the rectangle length for obtaining is：Point2.x-
Point1.x；Highly it is：Point2.y-Point1.y.
8) compare the rectangle recorded in template to grow tall, if error is within 8 pixels, then it is assumed that recognize successfully, return
The coordinate of fitted rectangle and grow tall；If the error of length or height wherein has one to exceed 8 pixels, next step is carried out；
9) next candidate line sections are taken out in the line segment aggregate of top, is fitted again.
10) when top line segment aggregate is all completed and does not find suitable rectangle, by the line segment of the direction
Index resets, and the authority for changing line segment is moved to next direction by the order according to " upper, left, down, right ".
11) after all line segment aggregates whole test fitting on four direction is finished, if suitable rectangle is not found
Output " rectangle does not find ".
Above-described embodiment is the present invention preferably embodiment, but embodiments of the present invention not by above-described embodiment
Limit, other any Spirit Essences without departing from the present invention and the change, modification, replacement made under principle, combine, simplification,
Equivalent substitute mode is should be, is included within protection scope of the present invention.
Claims (6)
1. in a kind of paper without anchor point pending paper rectangle posting recognition methods, it is characterised in that：
1) paper is read in, carries out binaryzation, correction and process；
2) rectangular area information of template record is read, with " largest contours searching " method the largest contours in region is obtained, used
Minimum rectangle is fitted to profile；
3) judge whether the rectangle size for fitting is consistent with the rectangle size of record, method is one error range of setting,
The length of two rectangles and the error of height are respectively compared whether in setting range；If in setting range, this is fitted
The rectangle for coming is the rectangle for needing identification；If not in setting range, carrying out step 4)；
4) line segment of area information is detected with Hough transformation, and the line segment to detecting is pre-processed, including classification, conjunction
And；
5) sequence is weighted to pretreated line segment；After being calculated the weights of every line segment, according to weights from big to small
Line segment is ranked up；
6) out, remaining gives up the line segments extraction that weights are located at into front three；
7) respectively take out a line segment from four direction up and down to be fitted, the method for fitting is：Line segment is obtained with left line
The intersection point Point2 of the intersection point Point1 of section, lower line segment and right line segment, being fitted the rectangle length for obtaining is：Point2.x-
Point1.x；Highly it is：Point2.y-Point1.y；
8) compare the rectangle recorded in template to grow tall, if error is within setting range, then it is assumed that recognize successfully, return fitting
The coordinate of rectangle and grow tall, it is the rectangle for needing identification that this fits the rectangle for coming；If the error of length or height wherein has
One exceedes setting range, then carry out next step；
9) next candidate line sections are taken out in the line segment aggregate of top, is fitted again；
10) when top line segment aggregate is all completed and does not find suitable rectangle, by the line index of the direction
Reset, and the authority for changing line segment is moved to next direction by the order according to " upper, left, down, right "；
11) after all line segment aggregates whole test fitting on four direction is finished, export if suitable rectangle is not found
" rectangle does not find ".
2. method according to claim 1, it is characterised in that the step 5) in line segment weights miss including segment positions
Difference weights and line segment length error weights；
Segment positions error weight computing formula is：
Site error weights=0.8^abs (current line segment sequence number records line segment sequence number).
Line segment length error weight computing formula is：
Error in length weights=(line segment length minimum line segment length)/(maximum line segment length minimum line segment length).
3. in a kind of paper without anchor point blank template paper selection rectangle method, it is characterised in that：
1) binary conversion treatment is carried out by scanning the original paper image for obtaining；
2) correction process is carried out to it；
3) after the image after being corrected, three maximum rectangles of area on paper are found out with " searching largest contours " method, and is remembered
Record its position；
4) rectangle is carried out once " judgement of rectangular line overlap ", it is ensured that largest contours are searched out going out with fitting a straight line for rectangle
The rectangle size come is consistent.
4. method according to claim 3, it is characterised in that the step 2) in correction process to be specially and search out figure
The maximum profile of area, is then surrounded profile with the rectangle of a minimum as in, calculates the angle of rectangle, draws inclining for paper
Rake angle, then paper is rectified a deviation.
5. method according to claim 3, it is characterised in that the step 4) " judgement of rectangular line overlap " method is as follows：
1) the pixel width of rectangle frame is determined, general 3～6 pixels are rule of thumb set to 5；
2) from positioning region from top to bottom transversal scanning image, 5 pixels are scanned successively downwards, if wherein there is black picture element
Point, then stop scanning, and adds 1 by total pixel SumPix of the row；
3) after a line is scanned through, the size of SumPix and positioning area field width is compared, if SumPix is more than positioning area field width
80%, then it is assumed that the row is likely to be the length of side of rectangle, and stops scanning；Otherwise, then by the row pixel whole zero setting；
4) from upper and lower, left and right four direction it is purged operation to image successively, filters off noise as much as possible；
5) straight-line detection is carried out with Hough transformation function to the rectangular region image after removing is processed, obtains the region
Line segment information；
6) line segment to detecting carries out after-treatment；
7) serial number information of rectangle line segment on four direction is recorded.
6. method according to claim 5, it is characterised in that the after-treatment step is as follows：
1) line segment is divided into four classes, one is the line segment on the vertical left side, and two is the line segment on vertical the right, and three is the line above level
Section, four is the line segment of horizontal bottom, sorting technique：The angle of line segment is calculated according to line segment coordinate, then compares line segment coordinate and area
The magnitude relationship at domain center, draws the classification of line segment；
2) line segment is merged：Two lines section is taken out from sorted line segment, their distances on respective direction are calculated, it is vertical
Line segment distance is the absolute value for subtracting each other of y-axis coordinate, and the line segment distance of level is the absolute value for subtracting each other of x-axis coordinate, compare away from
Whether from 5 pixels, if then merging in two straight lines, the method for merging is the coordinate for taking the respective end points of two straight lines；
3) line segment after merging is sorted, because to record index number of the rectangular shaped rim line segment in all line segments for identifying,
Therefore need that the line segment of diverse location is needed to make different sequences：
Top line segment：Sorted according to y-axis size；
Bottom line segment：Sort according to sections bottom positional distance size is found；
Left side line segment：Sorted according to x-axis size；
Right side line segment：Sort apart from size according to region right positions are found.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610938847.3A CN106650608B (en) | 2016-10-31 | 2016-10-31 | The recognition methods of rectangle posting in a kind of no anchor point paper |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610938847.3A CN106650608B (en) | 2016-10-31 | 2016-10-31 | The recognition methods of rectangle posting in a kind of no anchor point paper |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106650608A true CN106650608A (en) | 2017-05-10 |
| CN106650608B CN106650608B (en) | 2019-11-26 |
Family
ID=58820992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610938847.3A Expired - Fee Related CN106650608B (en) | 2016-10-31 | 2016-10-31 | The recognition methods of rectangle posting in a kind of no anchor point paper |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650608B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107423738A (en) * | 2017-08-02 | 2017-12-01 | 广东工业大学 | A kind of paper topic localization method and device based on template matches |
| CN107506746A (en) * | 2017-09-08 | 2017-12-22 | 电子科技大学中山学院 | Locating point-free image identification method and system for intelligent marking system |
| CN109272569A (en) * | 2018-08-03 | 2019-01-25 | 广东工业大学 | A kind of method that autocad builds X-Y scheme rapidly extracting and generates floor contour line |
| CN109977910A (en) * | 2019-04-04 | 2019-07-05 | 厦门商集网络科技有限责任公司 | Bill method for rapidly positioning and its system based on colored line segment |
| CN111626280A (en) * | 2020-04-13 | 2020-09-04 | 北京邮电大学 | Method and device for identifying answer sheet without positioning point |
| CN112163529A (en) * | 2020-09-30 | 2021-01-01 | 珠海读书郎网络教育有限公司 | System and method for uniformly dividing test paper |
| CN112200058A (en) * | 2020-09-30 | 2021-01-08 | 珠海读书郎网络教育有限公司 | System and method for intelligently correcting auxiliary data |
| CN112733855A (en) * | 2020-12-30 | 2021-04-30 | 科大讯飞股份有限公司 | Table structuring method, table recovery equipment and device with storage function |
| CN117409428A (en) * | 2023-12-13 | 2024-01-16 | 南昌理工学院 | Test paper information processing method, system, computer and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103793712A (en) * | 2014-02-19 | 2014-05-14 | 华中科技大学 | Image recognition method and system based on edge geometric features |
| CN104881663A (en) * | 2015-05-13 | 2015-09-02 | 京北方信息技术股份有限公司 | Method and device for discriminating selected result of check box |
-
2016
- 2016-10-31 CN CN201610938847.3A patent/CN106650608B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103793712A (en) * | 2014-02-19 | 2014-05-14 | 华中科技大学 | Image recognition method and system based on edge geometric features |
| CN104881663A (en) * | 2015-05-13 | 2015-09-02 | 京北方信息技术股份有限公司 | Method and device for discriminating selected result of check box |
Non-Patent Citations (4)
| Title |
|---|
| CL´AUDIO ROSITO JUNG 等: "Rectangle Detection based on a Windowed Hough Transform", 《PROCEEDINGS OF THE XVII BRAZILIAN SYMPOSIUM ON COMPUTER GRAPHICS AND IMAGE PROCESSING》 * |
| ZI-QIANG LI: "Generalized Hough Transform: Fast Detection for Hybrid Multi-Circle and Multi-Rectangle", 《PROCEEDINGS OF THE 6TH WORLD CONGRESS ON INTELLIGENT CONTROL AND AUTOMATION》 * |
| 宋黎明 等: "基于机器视觉的LED芯片检测方法", 《图形、图像与多媒体》 * |
| 王博凯 等: "基于感兴趣区域的快速虹膜定位方法", 《天津理工大学学报》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107423738B (en) * | 2017-08-02 | 2020-10-23 | 广东工业大学 | Test paper subject positioning method and device based on template matching |
| CN107423738A (en) * | 2017-08-02 | 2017-12-01 | 广东工业大学 | A kind of paper topic localization method and device based on template matches |
| CN107506746A (en) * | 2017-09-08 | 2017-12-22 | 电子科技大学中山学院 | Locating point-free image identification method and system for intelligent marking system |
| CN109272569A (en) * | 2018-08-03 | 2019-01-25 | 广东工业大学 | A kind of method that autocad builds X-Y scheme rapidly extracting and generates floor contour line |
| CN109272569B (en) * | 2018-08-03 | 2023-07-11 | 广东工业大学 | Method for quickly extracting and generating floor contour lines of autocad building two-dimensional map |
| CN109977910B (en) * | 2019-04-04 | 2021-08-20 | 厦门商集网络科技有限责任公司 | Rapid bill positioning method and system based on color line segments |
| CN109977910A (en) * | 2019-04-04 | 2019-07-05 | 厦门商集网络科技有限责任公司 | Bill method for rapidly positioning and its system based on colored line segment |
| CN111626280A (en) * | 2020-04-13 | 2020-09-04 | 北京邮电大学 | Method and device for identifying answer sheet without positioning point |
| CN111626280B (en) * | 2020-04-13 | 2021-09-07 | 北京邮电大学 | Method and device for identifying answer sheet without positioning point |
| CN112163529A (en) * | 2020-09-30 | 2021-01-01 | 珠海读书郎网络教育有限公司 | System and method for uniformly dividing test paper |
| CN112200058A (en) * | 2020-09-30 | 2021-01-08 | 珠海读书郎网络教育有限公司 | System and method for intelligently correcting auxiliary data |
| CN112733855A (en) * | 2020-12-30 | 2021-04-30 | 科大讯飞股份有限公司 | Table structuring method, table recovery equipment and device with storage function |
| CN112733855B (en) * | 2020-12-30 | 2024-04-09 | 科大讯飞股份有限公司 | Table structuring method, table recovering device and device with storage function |
| CN117409428A (en) * | 2023-12-13 | 2024-01-16 | 南昌理工学院 | Test paper information processing method, system, computer and storage medium |
| CN117409428B (en) * | 2023-12-13 | 2024-03-01 | 南昌理工学院 | Test paper information processing method, system, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106650608B (en) | 2019-11-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650608A (en) | Identification method for rectangle locating frame in test paper without locating points | |
| US4998285A (en) | Character recognition apparatus | |
| CN108596066B (en) | Character recognition method based on convolutional neural network | |
| US4028674A (en) | Automated signature verification system | |
| CN101859382B (en) | License plate detection and identification method based on maximum stable extremal region | |
| CN101923741B (en) | Paper currency number identification method based on currency detector | |
| CN104778470B (en) | Text detection based on component tree and Hough forest and recognition methods | |
| CN106529532A (en) | License plate identification system based on integral feature channels and gray projection | |
| CN104794491A (en) | Fuzzy clustering steel plate surface defect detection method based on pre classification | |
| CN105243730A (en) | Banknote identification method and system | |
| Diem et al. | Text line detection for heterogeneous documents | |
| CN102902967A (en) | Method for positioning iris and pupil based on eye structure classification | |
| CN111091124B (en) | Spine character recognition method | |
| CN109741314A (en) | A kind of visible detection method and system of part | |
| CN107392141A (en) | A kind of airport extracting method based on conspicuousness detection and LSD straight-line detections | |
| CN103310211A (en) | Filling mark recognition method based on image processing | |
| CN107452035B (en) | Method and apparatus for analyzing lane line image and computer readable medium thereof | |
| CN108509950B (en) | Railway contact net support number plate detection and identification method based on probability feature weighted fusion | |
| CN107506746A (en) | Locating point-free image identification method and system for intelligent marking system | |
| CN106815830B (en) | Image defect detection method | |
| CN106778717A (en) | A kind of test and appraisal table recognition methods based on image recognition and k nearest neighbor | |
| CN104680154A (en) | Identity recognition method based on fusion of face characteristic and palm print characteristic | |
| CN103258218A (en) | Matte detection frame generation method and device and defect detection method and device | |
| CN107463939A (en) | A kind of image key line detection method | |
| CN112597868A (en) | Test paper identification and correction method based on no positioning point |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20191126 Termination date: 20201031 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |